ascend-moe-optimizer-trace-analyzer

在用户提供 Chrome/Perfetto trace.json、或排查 Ascend 上 MoE/FusedDeepMoe 等算子性能时使用。按 phase、category、core group、tid 统计耗时、overlap、bubble，输出 CSV、Markdown 报告与确定性诊断；可选外部 LLM 扩写分析。默认 phase 映射面向 UMDK FusedDeepMoe，其它 trace 需替换或扩展 config/phase_map.yaml。

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

65%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A detailed, actionable skill body with strong executable commands and domain-specific definitions, weakened by length/repetition and implicit validation checkpoints, and by keeping reference-grade detail inline in a single file.

Suggestions

Tighten recurring content: consolidate the repeated default-config/FusedDeepMoe caveat and script-location explanations into single sections to reduce length and redundancy.

Add an explicit validation/feedback loop (e.g., check that expected CSVs/report.md were generated and non-empty, with a retry step) to raise workflow clarity.

Move reference-grade detail such as the full output-file catalog, default category list, and core-group/metric definitions into separate referenced files (e.g., OUTPUTS.md, METRICS.md) and link them from a concise overview.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and assumes competence, explaining domain-specific semantics Claude would not know (phase maps, core groups, union_us/overlap metrics); but ~210 lines with recurring repetition of the default-config/FusedDeepMoe caveat and script-location details could be tightened.	2 / 3
Actionability	Provides fully executable commands ('python3 app.py --trace ... --phase-map ... --output-dir ...'), a parameter table with defaults, concrete input-format specs (ph == 'X', B/E pairing), and a complete output-file catalog — copy-paste ready with placeholder substitution explained.	3 / 3
Workflow Clarity	Sequenced context-resolution and run variants are present (trace-only vs source+operator branches, validation command), but explicit validate→fix→retry checkpoints are largely implicit and output-correctness verification is not spelled out.	2 / 3
Progressive Disclosure	Organized into clear labeled sections, but it is a single monolithic file with no bundle files; reference-like detail (full output-file catalog, default category list, core-group internals, metric formulas) is inline rather than split into one-level-deep referenced files.	2 / 3
	Total	9 / 12 Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description that clearly states both what the skill does and when to use it, with natural trigger terms and a distinct niche. No first/second-person voice issues.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — '按 phase、category、core group、tid 统计耗时、overlap、bubble' and '输出 CSV、Markdown 报告与确定性诊断' — rather than vague language.	3 / 3
Completeness	Explicitly answers both what (analyze trace by phase/category/core/tid, output CSV/Markdown/diagnosis) and when ('当用户提供 Chrome/Perfetto trace.json、或排查 Ascend 上 MoE/FusedDeepMoe 等算子性能时使用').	3 / 3
Trigger Term Quality	Contains natural terms a user would say — 'Chrome/Perfetto trace.json' and '排查 Ascend 上 MoE/FusedDeepMoe 等算子性能' — with good coverage of relevant variations.	3 / 3
Distinctiveness Conflict Risk	Targets a clear niche — Ascend MoE/FusedDeepMoe Chrome/Perfetto trace analysis — with distinct triggers unlikely to match unrelated skills.	3 / 3
	Total	12 / 12 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: XiaoLuoLYG/GOD
Path: jiuwenclaw/jiuwenclaw/resources/agent/jiuwenclaw_workspace/skills/ascend-moe-optimizer-trace-analyzer/SKILL.md
Commit: 4fd5cfc

Reviewed: 2 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.